Unison

stefan · 24 May 2005

Eman Resu wrote: Is it possible to create a unison module? Would this have to be inserted before or after the midi to poly module?

Eman 🙂

I'd say that it will be possible and it will probably be possible both with midi components and with voice management components, whenever these things are introduced in SM 🙂

Josh · 24 May 2005

Thanks for the info stefancrs 😀

skOre · 24 May 2005

Only thing I know about Unison is "makes it fat and warm". Has anybody got a better explaination (including how it should work) ?

unkargherth · 24 May 2005

skOre wrote: Only thing I know about Unison is "makes it fat and warm". Has anybody got a better explaination (including how it should work) ?

AS far as i know, Unison mode is a mode of working for POLYphonic synthetisers

Supose a poly synth has one osc ("poor man synth"). If it allows polyphony of Six notes. the in fact has at least 6 osc. When you set in "Unison mode" you become limited to polyfony of one note, "monophony", but each of six osc are used for that note, usually 5 of them slightly detuned from "master one". Thats why Synth in unison modes soudn "Fat" and "Warm"

Never heard of but this should be also affordable in other combinations of Osc number/polyphony allowed. For example. get a synth with 16 notes polyphony limit to 4 notes and use 4 osc per note. A bit complex, but posible

Also, same concept should be aplicable to other sections. If you have a poly filter, in fact, you have one filter PER NOTE, so should be posible to slightly modify the parameters of filters in UNISON mode too

And, looping the loop, why not do the same for ADSR's, which are clearly per note

skOre · 24 May 2005

So it is kinda the same thing as superwave, right?

Believe me, detuning on things like Stereo spread and PWM modulation is already a stunning effect 😉

Hope to get to work on my Superwave again soon (4 days weekend this week, so hoping to find some time for it...)

unkargherth · 24 May 2005

skOre wrote: So it is kinda the same thing as superwave, right?

Believe me, detuning on things like Stereo spread and PWM modulation is already a stunning effect 😉

Hope to get to work on my Superwave again soon (4 days weekend this week, so hoping to find some time for it...)

Well, similar in some way. problem is thatyou use SEVEN REAL POLY oscilators, while unisonwill use seven MONO oscilators. So superwave uses for the same, seven times more cpu. But Obviously you could play more than one note, while in unison you only have one

Also you're spreading lot's of things. The Unison's i've seen only spread the tuning (althought i've made some test in SM spredaing filter cutoff and resand sonds nice and different)

I supose that originally Unison was a clever idea to give "More punch for same hardware" when playing Leads or basses. On digital world is usually cheaper use more "standard" components and limit polyphony to one (as you could do on Superwaveonce we get polyphony management in SM)

bones · 25 May 2005

That's pretty accurate although most synths I've owned with unison have allowed for detune and depan. Most modern synths, certainly all of my SE plug-ins, Alesis Mocron and Novation K-Station, allow you to choose how many voices to use in unison so that you can fatten up a patch and be able to play with reduced polyphony.

BTW, where does the superwave thing come from and what precisely is it?

unkargherth · 25 May 2005

bones wrote: BTW, where does the superwave thing come from and what precisely is it?

It's A SM Synth in Developement by SKore ( search the forum to find it, i don't have link at hand now)

Its niciest feature is that beign a simple synth, produces reaonably goo sound by spreading most values 7 times ( seven slightly detuned pitch, seven slightly diferent pan values, etc)

bones · 25 May 2005

No, its gotta be older than that because that's what SuperwaveP8 uses. I think it came from some Roland synth originally but I'm not 100% sure.

bones · 25 May 2005

Here's a thought, please correct me if I'm talking out of my arse. Given that oscillators don't use any resources until you trigger them, it should be possible to have multiple, redundant oscillators for use in unison. That would allow you to add unison without affecting the polyphony of your instrument but should have no CPU penalty in single-voice operation, right? It would also be far more efficient than cloning the entire signal path for each unison voice.

unkargherth · 25 May 2005

Well. By now, bones. This is only theory. We have no control over polyphony - jorge and/or malc have said that it will be in the future - so this topic it's a kind of brainstorming on Unison issue.

One we can said "POLYPHONY is now N" beigtn N a suficiently small number we can get Lots of nice Things just by populating signal path whith anything we want. Traditional Unison usually has done this By reusing free Oscs for detuning, and as you said , panning. I think would be nice to have aditional oscs slightly detuned AND dephased, for example, or aditional filters with cutoff and resonace modulated by semi-chaotic LFO's

The whole thing depends of the kind of access SM give us to voice management. Will be a totally diferent to have the traditional " Set polyphony to N" or -as jorge proposed some days ago - a complete set of voice management modules ( he didn't explained clear, but i suspect something in the line of : "New Voice, Clear Voice, Start Vice/End Voce, Voice combiner, etc). This last case makes that we can have our own "Voice politics" for voice stealing and so on. A bit complex, but much more open

EDIT: Sorry for the rant. I'm not answering you above

On what's is used and not, AFAIK things work this way. You can have any number of osc, filters, effects, etc in a schematic. They don't waste CPU, but it also they won't waste memory. In fact they're not even linked if they'aren't on the signal path (look here for an clear example of this). So, if you do things correctly, you could have a lot of extra oscs, and trigger it only when you want

stefan · 25 May 2005

bones wrote: Here's a thought, please correct me if I'm talking out of my arse. Given that oscillators don't use any resources until you trigger them, it should be possible to have multiple, redundant oscillators for use in unison. That would allow you to add unison without affecting the polyphony of your instrument but should have no CPU penalty in single-voice operation, right? It would also be far more efficient than cloning the entire signal path for each unison voice.

Atm, we have no control over polyphony. But apart from that, about redundant oscillators etc, you are fully correct. You can have support for 16 detuned oscillators if you like, you'd only use cpu for the ones that are enabled in the signal path anyway.

JMH · 25 May 2005

Superwave as a term does indeed come from a Roland synth as Bones recalls. From the JP-8000 to be exact. Loved by tarnce people and hated by all tarnce haters 🙂

Afair (too lazy to verify this with Google), JP-8000 had seven detuned oscillators coming out of a single oscillator in supersaw mode. Other manufacturers followed, such as Novation with their dualsaw mode. The current king of the hill must be Virus TI, it does something like 96 oscillators per note if you use the superwave mode on several oscillators and layer things up - and the end result isn't necessarily a chaotic mess of sounds. And tarnce people will buy it, that's for sure :P

Usually, 'superwave' does what unison does, but inside a single oscillator - thus there are no possibilities for separate panning and filtering of the detuned voices. Hopefully in the future we'll see synths that implement both, a 'supersaw' or 'superwave' mode for the oscillators and a separate unison mode with enough controls.

Now that I remember... as unison was discussed in another thread. One of the friendliest and most useful implementations of unison so far seems to be in ImageLine's Sytrus. Something people should check out when thinking about doing unison stuff. I'll post to the unison thread about this, once I dig up some more information on the subject (it's been "discussed" on KVR recently in several threads, for example).

* edit * in another thread... shows you how tired I am :P Gotta get some coffee and then post the rest of the stuff I have in mind.

Regards,

JMH

Nu Audio Science · 25 May 2005

Stop all this tarnce crap, Your not at KVR now 🙂 Seriously though i ain't a huge trance fan or something but i really hate that tarnce shit, If somebody started calling my music truntblaesim or something i'd hunt em down 🙂

Doesn't Superwave originate on the JX8P , I seem to remember it but i have owned a JP8000 aswell so i could be wrong ?

bones · 26 May 2005

JMH wrote: Usually, 'superwave' does what unison does, but inside a single oscillator - thus there are no possibilities for separate panning and filtering of the detuned voices. Hopefully in the future we'll see synths that implement both, a 'supersaw' or 'superwave' mode for the oscillators and a separate unison mode with enough controls.

I think this kind of hybrid is the perfect answer, a great compromise between performance and CPU usage. Cloning the whole signal path shouldn't really have any greater effect than cloning and detuning/depanning just the oscillators, as far as I can tell.

Now that I remember... as unison was discussed in another thread. One of the friendliest and most useful implementations of unison so far seems to be in ImageLine's Sytrus. Something people should check out when thinking about doing unison stuff. I'll post to the unison thread about this, once I dig up some more information on the subject (it's been "discussed" on KVR recently in several threads, for example).

I never noticed, and unison is a subject very close to my heart. What forum were these discussions in? I'll have to track 'em down. Thanks for the info.

BTW, when you guys say there is no control over polyphony, does that mean that any VSTi we create will have unlimited polyphony? Dangerous stuff if that's the state of play.

JMH · 26 May 2005

Heya 🙂

I can't recall nor seem to find any of the threads on KVR easily - it's almost a nuisance to find anything on KVR nowadays... But one of the threads were it was briefly discussed was about "is Melohman Minimonsta eating too much CPU", where people expressed their opinions about Minimonsta being a CPU hog without realizing that it uses a 'proper' form of unison, in other words really layering up to 6 (or whatever it is on Minimonsta) slightly detuned notes. Naturally this eats a lot of power, as filters etc. all done for each individual layer.

Me, I'd like to see both 'superwave' and 'proper unison' featured on a modern synth. As 'superwave' normally does the detuned layers in the oscillator section of the synth, the detuned layers are processed by a single filter etc in the signal chain and thus sounds markedly different to unison. That's one reason I dig Virus TI, you can do either, both or any combination.

Regarding that bit about doing 'superwave' instead of 'unison' to cut down CPU usage... I'd have to disagree on this. BUT. It has a lot to do with the rest of the signal path. With certain kinds of filters, let's take FabFilter One as an example, having 'unison' might sound rather horrible, due to the filter having such a character, especially with high resonance. But it most likely would sound good to my ears if it had an 'superwave' oscillator and a bunch of detuned variations were fed to the filter as one. With a real analogue sounding filter, 'superwave' would sound really great, as such filters start to saturate heavily (depending on the input volume of course) in a pleasing manner. One of my favorite sound generating devices, Commodore 64's filters does an incredible saturated, or rather, distorted, sound when you feed more than one oscillator through the filter. (And I really hope that some day developers will nail down this phenomenon well enough... Minimonsta does something similar but not as sweet sounding or musically useful)

And sorry for the obligatory tarnce references =) It's just a habit that pops up when ever discussing things with other KVR people... (my opinion of trance as a genre is that it sucks - too many wannabe 'producers' trying to sound like each other - but there exists a bunch of stuff classified as trance I like, and I've even done a few tranceish things myself, weirdly enough first as parodies on how easy that stuff is to write and later on being satisfied with my results to a degree that I wanted to finish the tracks and be satisfied with the results)

And no, 'superwave' as a feature was first presented in JP-8000. JX-8P could do killer fat sounds though, thanks to the architecture having a few oscillators per voice and being able to layer those... one of those synths I've always wanted to have.

(Curious sidenote, I do a faint recollection that there was a synth ages ago that could play a detuned version of an oscillator without eating polyphony, just can't remember what it was... it could be taken as a forefather to this thing we call superwave nowadays...)

I'll try to dig up more unison stuff after I get some sleep. I should have a small document on my hard disk that I wrote a while ago, thinking about how I'd like to see unison realized in a synth (I talk to some devs frequently, pushing my requests and ideas to them). What Gol did with Sytrus' implementation is pretty much what I've always had in mind, though it could still have some more controls.

Regards,

JMH who still can't shake off the megaposting habit

bones · 26 May 2005

JMH wrote: Me, I'd like to see both 'superwave' and 'proper unison' featured on a modern synth. As 'superwave' normally does the detuned layers in the oscillator section of the synth, the detuned layers are processed by a single filter etc in the signal chain and thus sounds markedly different to unison.

That would depend on your filter architecture, I imagine. I certainly couldn't see that there would be a difference in anything I've made in SE, especially if you compensate for the increased level of multiple oscillators, but I could be wrong.

stefan · 26 May 2005

bones wrote: BTW, when you guys say there is no control over polyphony, does that mean that any VSTi we create will have unlimited polyphony? Dangerous stuff if that's the state of play.

Yep. But the voice handling will be rehauled, the way it works now is by "magic" inside the midi2poly converter. As soon as the magic is exposed and becomes logic, we will have control over it.